How do I tell that an AI is speaking on the phone?

Most reliably, because it tells you: a properly set-up AI introduces itself as an AI at the start of the call. If that's missing, pay attention to how the conversation flows. Watch for short, even pauses before every answer, very consistent speech, and answers that don't really engage with personal side remarks.

How do I recognize an AI voice?

By sound alone, it's hard today, since modern AI voices sound natural. More telling than the sound is the behavior: a consistent response delay, almost no spontaneous sounds like laughing or hesitating, and an intonation that stays even even on emotional topics. So don't rely on your ear alone; test the call with targeted questions.

Can I throw an AI off with unusual questions?

Often, yes. Ask something unrelated to the matter at hand, reference something made up, or ask it to rephrase something in other words. A human reacts flexibly, asks in confusion or laughs. A simple voice system dodges, ignores the question, or stubbornly returns to its script. It's not proof, but it's a strong signal.

What do I do if a familiar voice calls and asks for money?

Hang up and call the person yourself, at the number you already have saved. Voices can now be recreated from very little audio, so a familiar voice alone is no longer proof. Never give out one-time codes, passwords or PINs over the phone, and agree on a codeword within your family that only you know.

Are all AI phone calls scams?

No. Many businesses put an AI on the phone so callers don't end up on voicemail: it logs requests, answers recurring questions, and records callback requests. The difference from a scam is in the behavior: a legitimate AI call states the business and the reason for calling, never asks for money or data, and doesn't apply pressure.

Detect AI Calls: A Phone Checklist

Q: Does a legitimate AI identify itself as an AI on its own?

Yes. The EU AI Act requires that people be able to recognize when they're interacting with an AI system. Reputable providers implement this by having the assistant identify itself as an AI right at the start of the call. Anyone who presents you with a human on the phone while a machine is actually speaking is violating exactly this principle. That's the clearest warning sign there is.

Detect AI Calls: the phone checklist.

The most reliable way to detect an AI call is that the AI introduces itself: legitimate systems say at the start that an AI is speaking. If that's missing, how the conversation flows and targeted follow-up questions reveal more than the sound of the voice. The checklist below walks through the signs step by step.

By Flo Knies — AI Automation Manager (TÜV SÜD Akademie), three IHK qualifications, trained HVAC and plumbing tradesman and field service technician.

Published July 30, 2026

Why you shouldn't rely on your ear alone

A few years ago it was simple: artificial voices sounded tinny, halting, unmistakably like a machine. That's over. Modern speech models breathe audibly, pause naturally, and stress words the way a human would. If you want to be sure today, you need more than a good ear: a few repeatable checks that focus on behavior rather than sound.

Important upfront: an AI on the phone is not, by itself, a warning sign. Many businesses use one quite openly, for example as an AI answering service, so that calls don't end up on voicemail. What separates legitimate AI calls from scam tactics is sorted out in detail in the overview AI phone calls: what's behind them. Here, it's about the practical question: how do you notice?

The checklist: six checks during the call

Listen to the opening. Does the voice introduce itself as an AI, with a business and a reason for calling? Then you know where you stand. If it claims to be human and the call still feels mechanical, be cautious.
Watch the response rhythm. Voice systems need a moment to answer. What's revealing isn't the pause itself but its consistency: the same short delay before every answer, no matter whether the question was easy or hard.
Throw in an unrelated question. Interrupt with something off-topic, like a question about the weather or a made-up reference. Humans react flexibly or with confusion. Simple systems ignore the interruption or return to their script word for word.
Ask it to rephrase. A human easily rephrases things. A weak voice system repeats the same sentence again, almost word for word.
Check the request, not just the voice. Money, one-time codes, passwords, gift cards, combined with time pressure and a demand for secrecy: that combination is the real warning sign, whether a human or a machine is speaking.
When in doubt: hang up and call back yourself. Call the person or business at the number you already have, not the one the call came from. This one step defeats almost every phone scheme, including ones with a cloned voice.

Recognizing an AI voice: what to listen for

Even though your ear alone isn't enough, there are clues. Artificial voices often keep pace and volume noticeably constant throughout the entire call. Spontaneous sounds are largely missing: no real laughter, no hesitation, no catching a breath mid-sentence. And on emotional topics, the intonation stays oddly even where a human would speed up, get louder or quieter. Why that is gets explained in the foundational article on voice AI: the voice is generated from text, and what isn't in the text, you won't hear either.

Treat these signals as a reason for the checks above, not as a verdict. There are people who speak in a monotone, and AI voices that sound lively. The sound gives you the suspicion, the behavior gives you the certainty.

How a legitimate AI introduces itself

The counterpart to a scam call is transparency. The EU AI Act requires that people be able to recognize when they're interacting with an AI system; since August 2, 2026, that's binding (more on that in the article EU AI Act and AI phone assistants). Reputable providers implement this by having their AI phone assistant identify itself as an AI right at the start of the call, before getting into the caller's request. This is a general explanation and not legal advice for an individual case, but it's the path that builds trust.

Our assistant Steffi is an AI too: in the sample call on her page, she introduces herself as a virtual assistant in her very first sentence. What a call like that feels like overall is described in the article Talking to AI on the phone. And what's behind topics like call recording and data protection, legally, is sorted out in the article Is an AI phone assistant GDPR compliant?

If the call still seems suspicious

Rely on the simplest rule: no legitimate caller minds you hanging up and calling back. Anyone who tries to keep you on the phone, pressures you to hurry, or asks for secrecy has something to hide. For a scam that caused damage, the police are the right contact; for a suspicious call claiming to be from your bank, the bank itself, via its official number.